Would it be possible to allow additional data from the experiments to be recorded into the database? Similarly to how hyperopt allows arbitrary attachments. Those could be useful to record e.g. additional metrics, not used in optimisation but good to have otherwise (e.g. finding best hyperparameters based on val scores, but saving also test scores for later reporting. Or optimising accuracy while saving also confusion matrices etc).
Would it be possible to allow additional data from the experiments to be recorded into the database? Similarly to how hyperopt allows arbitrary attachments. Those could be useful to record e.g. additional metrics, not used in optimisation but good to have otherwise (e.g. finding best hyperparameters based on val scores, but saving also test scores for later reporting. Or optimising accuracy while saving also confusion matrices etc).